Grok 4.7 edited this entire video. Here’s how it did
I swapped xAI’s newest model in as my daily driver. It built a 3D time-volume app and rotoscoped my footage from one prompt.
I swapped xAI’s newest model in as my daily driver. It built a 3D time-volume app and rotoscoped my footage from one prompt.
Grok transcribes the video, Jev ranks every segment, and real-time decisions open up far more than clipping.
TypeSafe’s classifier sorted 1,000 of my emails at about 200 milliseconds each, and the bill was a few cents.
It keeps working for hours on hard tasks, but its obsession with tests can trap it. Here’s the workflow that fixes that.
Agent-written tests mirror the code instead of protecting the product, so I made them opt-in.
Solo auto-starts and restarts OpenCode and a Caddy proxy, so every device connects to the same server.
Randomize fonts, colors, icons, radius and the component library, then start a project from the result.
The alpha is framework- and provider-agnostic, with per-model types, isomorphic tools and headless chat components.
Select any element, type a change, and Claude Code or Cursor edits the code without you leaving the page.
Bun stays open source with the same team. Anthropic gets the runtime that already ships Claude Code.
A half-day build put live Waze reports, police alerts and a satellite map on the Tesla screen. Then it took off.